Post FlightGear 2020.2 LTS changes
James has pushed some configuration changes to next (SG+FG) to raise various required versions of things.
If you’re developing on ‘something fairly recent’, you shouldn’t see any difference except it would be a wise time to wipe your build dirs and re-CMake from clean. If you’re on a ‘somewhat older’ Linux, the baselines being Ubuntu 18.04 (their previous LTS) and CentOS 7 (which our Jenkins runs), you should also be fine except for GCC, where you will need a version supporting C++14: in practice this means GCC 5.x or higher. (Or use Clang…) For both of these distros there are officially supported ways to get newer GCCs (we’re actually using GCC 9 on Jenkins now..), so I don’t expect this to be a problem.
As part of this we now require CMake 3.10, Qt 5.9 and OSG 3.4.1, and James started removing various legacy pieces from next around this. There are plenty more cleanups that can be done based on this : especially on the CMake and C++ sides. Things like Boost -> Std stuff, Cmake targets-as-objects support (eg, target_compile_features can be used now).
Windows and macOS should be unaffected BTW - because on those platforms you're always running something you downloaded which is fairly up to date. If you choose to go and download an ancient version of CMake or Qt, on either platform that’s your problem to fix - but you’ll get a configuration error at least.
The idea of doing this /now/ is that the build deps won’t change again during the next two years, barring any surprises.
None of this affects the release branches of course: they continue to work with OSG 3.2, GCC 4.x, Qt 5.4 and CMake 3.0. And you shouldn’t see any behavioural changes because of this either.
Let’s see if reality corresponds to all the theory above :)
08/2020: James pushed a large overhaul of the Cmake build files for FlightGear:
- we now use modern Cmake style nearly everywhere, i.e targets-as-objects for an OOP-style build system
- transitive dependencies are handled automatically (or at least, that’s the idea…), which simplifies things in many places (eg, audio library handling)
- system vs built-in dependencies is handled vis ALIAS targets, so most other code doesn’t care. We link to a target which is an alias for either our built-in version or the system one (via an IMPORTED target)
- we should have better isolation from system headers / versions of things. 
Changes that are already on next
- Include multiplayer aircraft in recordings.
- Continuous recording to file without any loss of detail.
- Periodic generation of a recovery snapshot, allowing resumption after a flightgear crash.
See new File - Flight Recorder Control dialogue.
- Continuous recording to file is uncompressed to simplify replay; e.g. 100MB for an hour's recording near EDDF.
- A recovery snapshot is actually a single-frame continuous recording, and will be called <aircraft-type>-recovery.fgtape. Load in the usual way using the Flight Recorder Control dialogue or on the command line with --load-tape=<aircraft-type>-recovery. Then do Ctrl-R to see replay dialogue and click on My Controls. I think that some aircraft may have no or limited support for My Controls though?
Code clean-ups and changes
Collecting what's going to change : this will also be used to work out what manual or automatic migrations are required to keep aircraft working. It's expected that for the first few months of 'next', there will be higher than usual breakage, i.e regular might not be possible.
As a general guideline, the minimum system to develop / build the code on will be Ubuntu 18.04 LTS (Bionic) : the versions below correspond to what is available in a stock Bionic install.
- Make C++14 (possibly C++17 to facilitate VSG migration?  ) the minimum required version (will make it easier to continue replacing Boost items with std:: ones)  
- Make CMake 3.6 the minimum required version: this will allow simplifying a bunch of compatibility logic in the build files  
- Use CMake OBJECT_LIBRARY to improve how we collect sources together when building each sub-dir 
- Make Qt 5.9 the minimum for the launcher Done 
- Drop 32-bit windows support 
- Drop the pre-2017.x MultiPlayer message format - this will fix warnings from some aircraft about MP packet overflows 
- Drop the KLN-89 code, since it's unused and not maintained for a long time: Canvas and the regular GPS code can easily implement a working KLN-89 or similar equipment now 
- Switch to Compositor mode as the only rendering option MR #213 (flightgear), MR #78 (simgear) MR #200 (fgdata) 
- Drop Rembrandt support from C++ (really part of the above) 
- Drop the C++ NavDisplay: the Canvas version replaces it 
Carriers and AI
Richard has been working on carriers since April 2020; originally the intention was to release as part of the 2020.2 LTS however the changes have grown into a set of new features rather than bugfixes and are therefore more suited to a longer development and review process.
- Done Improved view support - so that the nearest tower will include carriers
- Done Changes to core code to support moving towers.
- Multiple (tower) views - permit selection of LSO, PLAT and Tower as part of the "Tower view"; selection mechanism TDB. Consider revisions to the existing view system to better support this.
- Done Extend XML to include definitions for FLOLS touchdown position, LSO view position, Tower View position, deck angle.
- Improve 3d models for IFLOLS to ensure that a ball call can be made.
- Improve logic for IFLOLS lights; not quite an LSO simulation more of an approximation of the definitive cases (e.g. waveoff)
- Pending Animate the lineup lights on the stern (of Nimitz class).
- Improved support for Precision Approach Landing System (PALS) (AN/SPN-46); to support lineup deviation as well as glideslope (ball) deviation.
- Change approach deviations to use new aircraft reference point rather than eyepoint; it is probably more correct to use the eyepoint when in the cockpit view - but for external views this can be slightly inaccurate when outside the aircraft to completely wrong when in tower/LSO view. The reference point is taken as the aircraft position with an optional model defined x,y,z offset. This allows ball tuning on a per aircraft basis.
- Pending Improved support for AI logic to permit XML definition of e.g. LSO logic. Possibly a version of the autopilot / state machine that can be included as "logic blocks"
- Pending Review 3d model usage for all Nimitz class and either add LOD selection (low, medium, high)
- Pending Possible extra visual aids for Case 1 and Case 2 recovery training (e.g. a set of markers to fly through)
- Pending Better integration of carriers with JSBSim to permit catapult hookup
- Pending Simulate carrier operations Flight Deck Personnel, e.g. yellow shirts.
- Pending Animation of arrestor wires
- Pending Add some sort of trap grading (approach/touchdown plot)
see https://www.cnatra.navy.mil/local/docs/pat-pubs/P-816.pdf for T-45 carrier operations
Possible additional items
|Note CompositeViewer integration would entail raising our required OSG version to 3.6 |
- OSG 3.6 has the text display issue , currently osgText doesn't work under the ALS pipeline.  one issue that seems to occur in a few places is missing text elements on cockpit devices ... There are issues with osgText not working in OSG > 3.4.2. We need to find a migration solution for osgText that works with Compositor and newer OSG versions  ... another option was creating a Canvas for each text animation, but creating a FBO for every small piece of text doesn't sound very performance-friendly.  We need to find some way to approximate with the old OSG text did
- Remove the 2D panel code in favour of Canvas (this requires completing some work to load 2D panel elements as Canvas, from XML)  (as of 06/2020, Gaétan Allaert is still working on the replacement of the 2D panel by canvas ) 
- Remove the C++ HUD in favour of Canvas-based version: this requires some kind of migration script or framework, so we have at least the default UFO HUD available 
- Review Tim's original suggestion: Another [Canvas] optimization is to use a pre-render camera that isn't in the scene graph to do the Canvas rendering. Otherwise the Canvas textures are rendered at least twice, in the near and far cameras.  Tim Moore’s canvas-rendering-on-a-pre-camera fix, again should give us a nice perf win for Canvas-intensive acft, since within each camera pass (far camera, near camera), all the passes of a technique are run. Also, Mathias stated once: that it would be good to be able to specify a completely different scenegraph in some subcameras, i.e. for having panel like instruments on an additional screen/display for example.
- https://sourceforge.net/p/flightgear/flightgear/ci/next/tree/src/Viewer/renderer_compositor.cxx#l927 (compositor renderer)
- https://sourceforge.net/p/flightgear/flightgear/ci/next/tree/src/Viewer/renderer_legacy.cxx#l1842 (legacy renderer)
- As per James' comments: Investigate adapting the WindowBuilder to allow OSG windows to be created dynamically for features like Stuart's FG1000 to be shown in a separate window (canvas placement) 
- Review/integrate experimental C++ patches for new Canvas elements floating around on the forum, namely:
2do: list outdated / deprecated functions for removal (I remember one comment saying: remove after FG 3.0 or something)
- Clean up the GPS code with respect to FG_210_COMPAT :
property tree clean-up
remove /yasim/* "new" path is /fdm/yasim (since version 201x.?)